Yes, I agree, and thus the temporizing ???? on the 'anti'. Sadly, this class stack here is resisting arrest. The default ctor initializes. What we want is a default ctor that doesn't init, and perhaps a lazy init later if nothing else wanders along.
On Tue, Mar 25, 2008 at 6:01 PM, Glen Mazza <[EMAIL PROTECTED]> wrote: > I wouldn't necessarily consider it an anti-pattern. Constructor > injection is good when you need to document/demonstrate that a class > must *not* be instantiated if it does not have those certain properties > available. Of course, that can be rolled back to just using setter > injection in areas where performance is crucial. > > Glen > > Am Dienstag, den 25.03.2008, 16:45 -0400 schrieb Benson Margulies: > > I'm not entirely sure that I believe it, but it appears that the > following > > bit of spring from cxf.xml is disproportionately expensive. Why? Because > it > > uses a contructor with arguments. Why? Because spring has to think much > > harder to decide on a constructor than a set of set methods. > > > > So... I'm going to try adding set methods, a no-args constructor, and an > > init-method, and see what I see. > > > > If this works, I'll be on the hunt for other cases of the same > (anti????) > > pattern. > > > > > > <bean id="org.apache.cxf.resource.ResourceManager" class=" > > org.apache.cxf.bus.resource.ResourceManagerImpl"> > > <constructor-arg> > > <list> > > <bean class="org.apache.cxf.resource.ClasspathResolver > "/> > > <bean class="org.apache.cxf.resource.ClassLoaderResolver > "/> > > <bean class=" > > org.apache.cxf.bus.spring.BusApplicationContextResourceResolver"/> > > </list> > > </constructor-arg> > > <property name="bus" ref="cxf"/> > > </bean> > >
